摘要
Urban air pollution remains a persistent challenge for global cities. The Local Climate Zone (LCZ) framework, which classifies urban environments based on land cover, physical geometry, and thermal properties, has shown significant potential for air pollution research. However, few studies have quantitatively explored the relationship between air pollution and LCZs. To address this gap, the study combines ultra-high-resolution pollutant concentration data (PM 2.5 , PM 10 , NO 2 , NO x ; grid size 20m × 20m) with LCZs (grid size 100m × 100m) and applies interpretable machine learning models. The Extreme Gradient Boosting algorithm, combined with the SHapley Additive exPlanations is used. Findings from Greater London show that the trained machine learning model incorporating LCZ categories, spatial coordinates, and buffer features provides strong performance, with testing R² of 0.89 for PM 2.5 , 0.85 for PM 10 , 0.90 for NO 2 , and 0.82 for NO x . The most polluted LCZs (the three highest deviations above the overall average) are concentrated within only 3.82% of Greater London. These include LCZ1 (Compact highrise), LCZ2 (Compact midrise), LCZ10 (Heavy industry), and LCZE (Bare rock or paved surfaces), where maximum pollutant concentrations are up to 68% higher than the overall average. However, the LCZs ranked as having the strongest overall impact on air pollution do not always correspond to those with the highest pollution levels. Feature importance analysis shows that LCZ5 (open midrise) has a significant effect on air pollution in Greater London. This is partly due to dispersion of pollutants from highly polluted areas (particularly LCZ1, LCZ2 and LCZ10 in Inner London) towards Outer Greater London, combined with emissions within LCZ5, leading to a cumulative impact on overall pollution levels. Overall, the results show the LCZ framework, despite some limitations, is a valuable approach for quantifying air pollution research, providing insights not only into the spatial distribution of high-pollution areas but also into the key factors driving urban air pollution patterns.